Google’s faked Gemini demo recreated with GPT-4
Google’s highly anticipated AI offering, Gemini, fell short of expectations last week after it was revealed that the impressive demo was, in fact, faked. However, one expert, Greg Sadetsky, has taken it upon himself to showcase the true capabilities of Gemini by recreating the demo using GPT-4. What sets Sadetsky’s recreation apart is the absence of any editing tricks used to enhance Google’s demo. While GPT-4 does deliver on what Gemini claimed to achieve, it does so at a slower pace and with less precision.
The disappointing turn of events surrounding Gemini has caught Google fans off guard. They had long been singing the praises of Google’s internal capabilities, often comparing them to rival AI organizations like OpenAI. The stumble at the starting line for Gemini comes as an unexpected setback for these dedicated supporters.
Interestingly, on another front, it has come to light that Google has recently disabled RSS on its blogs. Although RSS has lost its mainstream relevance, Google’s decision to keep it alive until now was seen by many as nothing more than tokenism or a mere attempt at displaying transparency. However, it appears that Google’s internal culture, characterized by a relentless focus on new projects and a disregard for maintenance, has finally prevailed. The demise of RSS at Google may be connected to a new development that caught the company’s attention, rendering RSS obsolete in their eyes.
The question arises: If working on established projects is considered unfashionable at Google and maintaining ventures that are not profitable is deemed undesirable, has Google’s AI plans already triggered the company’s own immune response? Google’s AI initiatives are intertwined with DeepMind, a British startup it acquired in 2014. It is possible that DeepMind’s integration was not successful, or that its approach to AI clashed with Google’s ever-evolving focus. Perhaps it is amusing to imagine that the internal psychology of Google employees triggers a premature inclination to move on to the next big thing as soon as they encounter Demis Hassabis, the founder of DeepMind.
In the world of rapidly advancing technology, even industry giants like Google are not immune to setbacks and surprises. While Gemini may have stumbled with its misleading demo, the recreation using GPT-4 shows that the capabilities are there, albeit in a slower and less refined manner. This turn of events brings into question Google’s internal culture and the implications it may have on its AI endeavors. As the journey of Google’s AI projects continues, it remains to be seen how they will navigate the delicate balance between innovation and preservation.
